Credit Ratings and Capital Structure : what is the relationship and has this changed over time?

(1)

Credit Ratings and Capital Structure: what is the relationship

and has this changed over time?

Abstract

Current capital structure theories don’t incorporate credit ratings, while it is increasingly argued that credit ratings matter when it comes to capital structure. This paper contributes to this debate by investigating the effect of a credit rating change on the issuance of debt. Using quarterly data of the S&P500, a regression shows that there indeed is a positive effect of an upgrade on debt issuance, while a downgrade causes a negative effect. This negative effect tends to be larger when the

downgrade results in a speculative graded credit rating. Overall, firms seem to be deleveraging since the start of the financial crisis. When including interaction variables and comparing different periods, it is shown that the effect of a downgrade has increased since the start of the financial crisis of 2008. The maturity of debt issued after a credit rating change is also studied in this paper, and results indicate that upgrades and downgrades cause more issuance of short-term and long-term debt respectively.

Keywords: Credit Rating, Capital Structure JEL Classification: G24, G32

Bachelor Thesis

Name: Rik Stapersma

Student number: 10385703

Programme: Economics & Business Track: Finance & Organization

Supervisor: J. Lemmen

(2)

2

Statement of Originality

This document is written by Rik Stapersma who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

Basic capital structure theory of Modigliani and Miller states that in perfect markets, capital structure shouldn’t matter when it comes to company value. However, we do not observe perfect markets in reality thus capital structure does matter. A factor that might be relevant when choosing the optimal capital structure is the credit rating of the firm. This research will try to investigate the relationship between credit rating and capital structure.

There have been previous studies on the relationship between credit rating and capital structure, but only recently this has been studied empirically. Those empirical studies have mainly focused on specific samples over a certain period of time, so it might be the case that these studies aren’t externally valid. Another problem with those studies is that they seem to contradict each other, so there isn’t a conclusive answer. I will look at the effect of a change in credit rating on capital structure, whereas previous research mainly looked at the effect that the current credit rating has on capital structure. The reason I focus on a change in credit rating is that it might be the case that managers only act on the capital structure when they get such a signal. A credit rating downgrade is considered a negative signal in the market and managers might want to get back to their original credit rating as soon as

possible. The recent financial crisis might have changed the way managers think of this, since credit rating agencies have been under criticism for their role in the crisis. It is possible that this also changed the way managers act upon credit rating changes, so this is something I would like to include in my research. Therefore, the research question is the following: what is the effect of a credit rating change on the capital structure of companies in the United States, and did the recent financial crisis change this?

I will use the companies included in the S&P500 index to estimate the effect of a credit rating change on capital structure during the period 2002-2014. To make sure the credit rating change corresponds to the effect on capital structure I will use quarterly data. First I will test the overall effect of credit ratings on capital structure, then I will look at the effect a change in credit rating has. Finally I will include interaction variables for different periods to estimate if the effect of a credit rating change differs over time. Results indicate that credit rating upgrades and downgrades have positive and negative effects on the issuance of debt

(5)

5 respectively. Furthermore indications are found that the effect of a downgrade has become larger since the start of the financial crisis. When investigating the effect of a credit rating change on debt maturity, results show that an upgrade causes firms to issue more short-term debt while a downgrade causes the issuance of relatively more long-short-term debt.

The rest of this paper is organized as follows. Section II will be used to summarize and discuss theories and previous findings on the relationship between credit ratings and capital structure. In Section III I will discuss the methodology used in this paper. Section IV gives the results from the empirical tests and discusses them. I conclude in section V.

II. Literature review

This section will start with a discussion of credit ratings and their role in the financial markets. I will also discuss in which ways this has changed over time, since this paper investigates a different sample period than previous literature on this subject. Next I will briefly discuss the most dominant theories in capital structure, since this can indicate where this research can be placed in current capital structure theories. Furthermore these theories give control variables used in the models in this paper. Finally I will discuss previous

literature on the relationship between capital structure and credit ratings.

a. Credit ratings

It is never completely certain that you will get the fixed income stream from an investment in a bond. There is a risk that the issuer might default on this obligation. A credit rating is a measure of this risk. There are three world leading companies that measure credit ratings, Standard and Poor’s, Moody’s and Fitch. They assign letter grades to corporate and

municipal bonds that gives information about the safety of the investment, the higher the grade the higher the creditworthiness of a company (see Appendix A for an overview of credit ratings). To assess this creditworthiness, the credit rating agencies need certain determinants. According to S&P’s their general criteria for creditworthiness are default likeliness, payment priority, recovery and credit stability. They also mention certain variables used to determine credit ratings, such as long- and short-term debt divided by total assets.

(6)

6 When determining credit ratings, agencies mainly look at the long term (Altman and Rijken, 2004).

Previous research has investigated the informativeness of credit ratings. Holthausen and Leftwich (1986) looked at the effect of credit rating changes on stock prices. Their main results are that they find no significant abnormal returns after an upgrade, but they do find significant negative abnormal returns after a downgrade. They argue that this is evidence that credit rating agencies provide information not already captured by the stock price. Dichev and Piotrovski (2001) study the same effect but they use long-run stock returns and a larger sample. Their results are consistent with the ones of Holthausen and Leftwich.

Another finding of theirs is that a credit rating downgrade is a strong predictor of future earnings deterioration. Boot et al. (2006) agree that credit rating agencies speed up the availability of information to the financial markets. This is all evidence that credit ratings do give information about firms that wasn’t previously available.

If credit ratings thus are informative about a company, it is then important to know what the costs (benefits) will be if a firm experiences a credit rating change. Direct costs of debt issuance, like commission fees or legal costs, will be higher with a lower credit rating

according to Lee et al. (1996). Hite and Warga (1997) find that a downgrade results in lower bond prices, which indicates a higher yield. The debt cost of capital thus increased because of the downgrade. This is also intuitive, since investors require a higher return for the increase in default risk. Kisgen (2006) argues that a lower credit rating negatively affects employee and customer relationships and access to commercial paper, while it may also cause discrete costs like bond repurchases. He also states that these effects will be larger for the change from investment-grade ratings to speculative-grade ratings. In addition to this, access to capital will be more difficult when downgraded to a speculative-grade rating, since certain types of investors aren’t allowed to hold these types of bonds. These findings

confirm that credit ratings play an important role in the corporate world.

However, the way credit ratings are assessed has changed over time. According to Baghai et al. (2014) credit rating agencies have become more conservative over time. The sample period they investigate is 1985-2009. They find that standards have become tighter and that

(7)

7 on average credit ratings declined by three notches holding firm characteristics constant. This is consistent with the findings of Blume et al. (1998), who study the same matter over the period 1978-1995. Baghai et al. (2014) claim that the increased conservatism is

unwarranted since they find a significant decline in defaults over their sample period. Another study of how ratings have changed over time comes from Becker and Milbourn (2011). They study the effect of increased competition on credit rating quality. For this they use the rise of Fitch as a credit rating agency. Their results indicate that the quality of S&P’s and Moody’s ratings declined when Fitch’s market share rose, because CRAs find reputation more important in less competitive industries. One of their measures of quality is the ability to predict defaults, so Becker and Milbourn’s results are similar to Baghai et al.’s in that credit ratings have become less informative about defaults.

Dimitrov et al. (2015) conduct a research that partly ties in with Becker and Milbourn’s. They study the impact of the Dodd-Frank act on credit ratings. The Dodd-Frank act, which passed in 2010, contains a series of reforms regarding credit ratings agencies. These reforms should encourage credit rating agencies to issue accurate ratings and to minimize material misstatements and fraud. Dimitrov et al. (2015) however find that following the passing of the Dodd-Frank act, credit ratings tend to give more false warnings than before, which means that a speculative graded issuer hasn’t defaulted within one year. They also find that bond markets respond less to credit rating downgrades than before, a result consistent with less accurate ratings. The link between this research and Becker and Milbourn’s is that they also find a strong reputational factor in that credit rating agencies tend to be more

conservative in less competitive industries. Cheng and Neamtiu (2009) study a similar event, the Sarbanes-Oxley act in 2002. Contrary to Dimitrov et al. they find an increase of credit rating accuracy and timeliness, as well as a decrease in credit rating volatility.

Since the financial crisis in 2008 credit rating agencies have been under scrutiny. I will not go into depth about the causes of the financial crisis, but I will comment briefly on the role of credit rating agencies. Scalet and Kelly (2012) summarize the events of the financial crisis regarding credit rating agencies. Since debt instruments are rated by CRAs on their default likeliness, it was an unexpected shock that a lot of these highly rated instruments actually did default in 2008. Investors relied on these ratings and were thus left to bear the costs.

(8)

8 They however could not take any legal action since a credit rating was legally just an opinion. Scalet and Kelly further mention some points of criticism that are often named when it comes to credit rating agencies. A first point of criticism is a conflict of interests. This arises because the CRAs are compensated by the companies whose bonds they rate, something studied by Stolper (2009) as well. Another point of criticism is a lack of transparency, since CRAs didn’t disclose the risks of investing in complex debt instruments leading up to the crisis. A third point of criticism is a lack of competence. They state that instruments that are too complex shouldn’t be rated by credit rating agencies. A final point of criticism that Scalet and Kelly mention is a lack of competition. According to them this makes it possible for the other three problems to arise. However, as mentioned earlier, increasing competition may also lead to less informative ratings (Becker and Milbourn, 2011). The factors mentioned here are reason to believe that overall trust in the informativeness of credit ratings has diminished.

b. Capital structure

There are two main theories of capital structure, the Trade-off theory and the Pecking Order theory. However, the first theory on capital structure is the famous Modigliani and Miller (1958) theorem that in perfect markets capital structure doesn’t matter. Since we do not observe perfect markets in reality this theorem needs to be adapted to suit the financial markets we do observe. These adaptations are made to include bankruptcy costs, taxes, agency costs and more.

The Trade-off theory incorporates some of these market imperfections. The main feature of this theory is that firms try to balance the costs and benefits of an additional dollar of debt, thus resulting in an optimal amount of leverage. In 1963 Modigliani and Miller added to their theorem the benefits of debt as a tax shield, while in 1973 Kraus and Litzenberger expanded this framework with the bankruptcy costs of debt. Since then it has been argued that

additional factors should be included in the model, such as the agency costs and benefits of debt (Jensen and Meckling, 1976). All these factors result in a target optimal level of debt. Some research has indeed shown that firms have a target level of leverage that they adjust

(9)

9 to (e.g. Leary and Roberts, 2005 or Harford et al., 2009). Flannery and Rangan (2006) also come to this result, and they argue that firms adjust to this target level relatively quick.

The Pecking Order theory was first introduced by Myers in 1984. He argues that when selecting financing methods, firms prefer retained earnings to debt, and debt to equity. The theory implies that debt will increase if an investment needs more capital than is internally generated, and will decrease if the investment needs less capital than internally generated. The motivation for this particular order is adverse selection according to Myers (1984) and Myers and Majluf (1984). If a firm issues equity, outside investors may wonder why they don’t issue debt and conclude that the company isn’t healthy enough to pay for the debt. The reason internal investment is at the top of the order is also intuitive, in that managers don’t like the costs and obligations of external finance. Myers (2001) also supports this order using agency theory arguments.

Fama and French (2002) test the two theories to decide which is a better predictor of debt levels and dividend payout. They find that both models are similarly good at predicting, however both also have some flaws. Nevertheless they conclude by stating that most shared predictions by the two models are supported by empirical evidence and that they can’t decide what’s the better theory of the two.

Using these theories, it is important for my research to identify factors that influence the amount of leverage. This way I can control for these characteristics. Rajan and Zingales (1995) identify 4 factors that they consider the main determinants of leverage, which are: tangibility, profits, firm size and market-to-book asset ratio. As a further contribution to this subject, Frank and Goyal (2009) studied the effect of firm characteristics on leverage of publicly traded American firms during the period 1950 to 2003. They expand the 4 factors found by Rajan and Zingales with industry median leverage and expected inflation as other determinants of leverage. Kayo and Kimuro (2011) study the same, however they use a data sample including 40 different countries. They also include industry specific factors and country specific factors in their research, but they find that the firm specific characteristics explain 78% of firm leverage. This is reason to include those characteristics in my research.

(10)

10 Another factor that might influence the capital structure of a firm is the interest rate.

Interest rates determine the cost of debt, so this potentially could have an impact on the capital structure. One large driver of the interest rate in the sample period of this paper is the Quantitative Easing program of the US Federal Reserve. The goal of this program was to stimulate the economy by large scale asset purchases. These asset purchases caused the Treasury interest rates as well as the corporate rates to drop according to Krishnamurthy and Vissing-Jorgensen (2011), which might also have had an impact on the capital structure of firms. Joyce et al. (2011) argue that this is indeed the case and find that there was a substantial increase of corporate bond issuance in 2009. Duca et al. (2014) support this finding and say that for advanced economies like the US the impact of QE on corporate bond issuance was concentrated at the beginning of 2009. These findings indicate that the

Treasury rate might be a factor that should be included in this research. In the methodology section I will discuss which control variables I will include and why.

c. The link between credit rating and capital structure

The main issue that is dealt with in this paper is the relation between credit rating and capital structure. Kisgen (2006) described this relation previously as the Credit Rating – Capital Structure Hypothesis, or CS-CR in short. He says that the CS-CR is different than financial distress arguments because CS-CR takes place at any credit rating, while financial distress probably only takes place at lower credit ratings. When looking at the Trade-off Theory, Kisgen argues that a credit rating change involves certain discrete costs and benefits and thus leads to a different optimal amount of leverage. He also incorporates the CS-CR in the Pecking Order Theory and argues that when making capital structure decisions, firms face a trade-off between the cost of issuing debt and the cost of a potential change in credit rating.

Kisgen (2006) furthermore contributes to this debate with an empirical research. He looks at the debt issuance of firms close to a credit rating upgrade or downgrade, arguing that these firms are less likely to issue debt since they will receive benefits (costs) from an upgrade (downgrade). His sample period is 1989-2001. The main findings of this research is that firms

(11)

11 close to a ratings change issue 1% less debt than other firms. Kemper and Rao (2013) find the same effect but they argue that the results are mostly driven by extremely low credit ratings. It is nevertheless still possible that credit ratings are an important factor in determining capital structure policy according to them. Kisgen expands on his research in 2009, looking at the net debt issuance after a credit rating change. He then finds that firms are less likely to issue new debt and to reduce equity, and more likely to reduce debt after a rating downgrade. An upgrade however has no significant effect.

Li et al. (2014) also studied the CS-CR hypothesis, but they mainly focused on the real estate sector by looking at REITs. While this sector has very specific characteristics, for example very high leverage, their results might still be interesting for the overall CS-CR debate. They look at both the ex-ante behavior of firms near a credit rating, as well as ex-post behavior after a credit rating change. What they find is similar to the results of Kisgen (2006) in that firms near a credit rating issue less debt. While they find no significant results on the effect of a credit rating change, they conclude that their results support the CS-CR hypothesis.

Other evidence that credit ratings matter when making capital structure decisions comes from Graham and Harvey (2001). They did a survey among CFOs of 392 US firms in different sectors. Their main findings regarding capital structure is that credit ratings are considered the second most important factor when determining debt policy. This is a surprising result given that interest tax savings and distress costs, frequently mentioned in capital structure theory, are of a far less concern in this survey. They also find that credit ratings are more important for large firms. Brounen et al. (2006) conducted a similar survey among CFOs in Europe where they find similar results regarding the importance of credit ratings in the capital structure decision, despite the fact that rating agencies are less active in continental Europe and that there are large institutional differences in Europe.

The level of a firm’s credit rating may not only be a determinant of capital structure, the fact that a firm has a credit rating might already influence this. This is what Faulkender and Petersen (2006) investigate. They find that firms with a credit rating have a significantly larger amount of debt. This is intuitive given that these firms have access to the public debt market whereas firms without a credit rating do not. Also, firms with credit ratings have

(12)

12 different characteristics, which may explain the difference in leverage. However, when controlling for these firms characteristics, Faulkender and Petersen still find a significant higher level of debt for firms with a credit rating. This is an interesting result since this indicates that different sources of capital (for example public vs. private) result in different capital structures. Mittoo and Zhang (2010) support this evidence when looking at Canadian firms.

I mentioned earlier that previous research found increasing conservatism of credit rating agencies. This also has an impact on capital structure according to Baghai et al. (2014). They use a model to predict credit ratings over the sample period of 1997-2009 and compare these with the actual ratings. What they find is that firms with a lower actual rating than predicted rating have a lower level of debt. This is another indication that firms tend to look at credit ratings when making capital structure decisions, however it is shown here that the credit ratings aren’t completely informative. It will be interesting to see if the impact of credit ratings on capital structure has changed given the criticism credit rating agencies have had in the past few years.

It also can be the case that firms want to change their credit rating before issuing debt, which is the subject of the research of Begley (2014). He argues that since credit rating agencies look at certain criteria (e.g. Debt/EBITDA ratio) when determining a credit rating, firms that are about to issue debt try to influence these criteria. The results he find support this argument, firms do tend to cut on R&D expenses for example to influence the criteria of credit rating agencies. While this research doesn’t look at the effect of credit ratings on capital structure, it does shows that credit ratings have an important effect on investment decisions.

A final factor that may be interesting to look at is the effect of credit ratings on the maturity of the issued debt. This is what Diamond (1991) investigated. Since short-term debt matures before the cash flows arrive, it must be refinanced and at different terms considering the credit rating at the time of refinancing. In contrast, long-term debt matches cash flows and future credit ratings are then thus less important according to Diamond. This results in that firms with high credit ratings tend to issue short-term debt, since their future outlooks on

(13)

13 credit ratings are positive and thus the future terms of the debt are positive as well. Firms with a somewhat lower credit rating will issue long-term debt however, because future refinance costs might be high if they used short-term debt. Finally, firms with low credit ratings are more likely to issue short term debt, since borrowers don’t want to issue long-term loans to them. Bali and Skinner (2006) also studied this empirically, and found that firms with higher credit ratings issue bonds with a shorter maturity. There however is no previous research on the effect of a credit rating change on the change in debt maturity. Overall can be concluded that credit rating matters when determining capital structure.

III. Methodology

In this chapter I will discuss the data used in this paper. Furthermore I will discuss the

models used to answer the research question, and the variables included in the models. I will use OLS regression to check whether a change in credit rating results in a change in capital structure. Since the capital structure might also influence the credit rating, it will be difficult to estimate the precise effect of a credit rating change. I must clearly distinguish the event and the control variables to capture the effect. To make sure endogeneity isn’t an issue, the change in capital structure will be lagged so that it is clear that this change only arises after the change in credit rating. This is the same way previous studies dealt with this endogeneity issue.

First I will run an OLS regression of the credit rating on the capital structure of companies. This is to see the overall effect of credit ratings on capital structure (1). Next I will include a dummy that indicates if a company has had a credit rating upgrade or downgrade (2). Finally I will include a dummy for the period before the start of the financial crisis in 2008, and interact this dummy with the credit rating change (3). The models I will estimate are the following:

𝑁𝑒𝑡𝐷𝐼𝑠𝑠 = 𝛽0+ 𝛽1𝐶𝑅 + 𝛽2𝐶 + 𝜀 (1)

(14)

14 𝑁𝑒𝑡𝐷𝐼𝑠𝑠 = 𝜃0+ 𝜃1𝐶𝑅𝑑𝑜𝑤𝑛 + 𝜃2𝐶𝑅𝑢𝑝 + 𝜃3𝐶𝑅𝑑𝑜𝑤𝑛 ∗ 𝑃𝑟𝑒𝐹𝐶 + 𝜃4𝐶𝑅𝑢𝑝 ∗ 𝑃𝑟𝑒𝐹𝐶 + 𝜃5𝐶 + 𝜀 (3)

The measure of capital structure I use here is NetDIss, which stands for Net Debt Issuance. The reason to use Net Debt Issuance is that it will indicate the relative amount of debt issued in the period after a credit rating change. NetDIss can be calculated the following way: 𝑁𝑒𝑡𝐷𝐼𝑠𝑠 = (∆𝐷 − ∆𝐸)/𝐴, where D is the level of long-term and short-term debt, E is the value of shareholder’s equity and A is the level of total assets. CR indicates the credit rating, CRdown is a dummy indicating if a firm’s credit rating has gone down, CRup is a similar dummy indicating an increase in credit rating, PreFC is a dummy indicating the period before the start of the financial crisis, C is a set of control variables and ε is the econometric error. The coefficients of interest are γ₁ and γ₂ in the second equation, since these will give information about the effect of a credit rating upgrade or downgrade on the capital

structure. In the third equation the coefficients of interest are θ₃ and θ₄, which will indicate if there’s a different effect between periods.

In addition to the models mentioned above, I will also use a model to estimate the effect a change in credit rating has on the maturity of the debt issued (4). The model used for this is the following:

𝑁𝑒𝑡𝑆ℎ𝑜𝑟𝑡𝐷𝐼𝑠𝑠 = 𝜑₀+ 𝜑₁𝐶𝑅𝑑𝑜𝑤𝑛 + 𝜑₂𝐶𝑅𝑢𝑝 + 𝜑₃𝐶 + 𝜀 (4)

CRdown, CRup, C and ε are defined the same way as above, NetShortDIss stands for the short-term debt issuance relative to the issuance of long-term debt. NetShortDIss can be calculated the following way: 𝑁𝑒𝑡𝑆ℎ𝑜𝑟𝑡𝐷𝐼𝑠𝑠 = (∆𝑆𝐷 − ∆𝐿𝐷)/𝐴, where ΔSD is the change in short-term debt, ΔLD is the issuance of long-term debt minus the reduction of long-term debt and A is the level of total assets. The coefficients of interest in this model are ϕ₁ and ϕ₂, which indicate the effect of a credit rating upgrade or downgrade on the maturity of the issued debt.

As control variables I will look at other factors that might influence the capital structure decision. These control variables come from literature previously mentioned. First I will use size as a control variable, and size will be measured by sales. This is known to be a good

(15)

15 predictor of firm size (see e.g. Frank and Goyal, 2009). I expect that size will have a positive effect on Net Debt Issuance, since larger firms find it easier to attract capital than smaller ones. According to Rajan and Zingales (1995) this is because these firms are better known in the market and have better reputations. Profitability is another control variable that I include, which is measured by the operating return on assets. The expected effect of

profitability is negative based on a Pecking Order Theory argument. Because firms with high profitability have more free cash flow left for investment they thus need less external financing. As a third control variable I will include the current debt/assets ratio, this might also be informative about net debt issuance. I expect the effect of the current debt/assets ratio to be positive, since firms with high current debt are more likely to issue new debt. A fourth control variable is the industry level of debt, which will control for certain industry characteristics. Other factors that might influence leverage, like R&D expenses or market-to-book asset ratio, aren’t included in the model because of unavailability of data. However, since previous research on the CS-CR hypothesis also excluded these factors, this isn’t a matter of concern. While it is intuitive to include the Treasury interest rate in the model, which can be seen as a benchmark to compare corporate rates with, this mostly isn’t done in previous research on the CS-CR hypothesis. To check for robustness I will include the 10-year Treasury rate however, as well as the extra control variables intangibility and depreciation, which hasn’t been previously done in empirical research but are known to be predictors of capital structure. This will be more extensively discussed in the results section. As done in previous research (e.g. Li et al., 2014) the control variables will be measured as a ratio of total assets, while for size the natural logarithm is used. The reason for this is that data for sales tend to be skewed.

The data I will use are the credit rating of Standard and Poor’s and the capital structure of American firms listed on the S&P500 as of December 31, 2014. I will use these firms because I would first like to investigate if I find a similar effect of credit ratings on the capital

structure as shown in previous literature. The main reason to use American data however is availability. To check for robustness I will also use an expanded dataset including both the S&P500 and the S&P MidCap 400. A drawback to this might be that the smaller firms of the S&P MidCap 400 might have less access to the public debt market. These firms are however still relatively big, so I assume that this won’t be a problem. Financial firms and Utility firms

(16)

16 will be deleted from the sample, since these firms are subject to special circumstances and it is common to leave these firms out of a data sample. The time period I choose to investigate is 2002-2014. I choose this period because this way I have roughly the same amount of data before and after the start of the financial crisis.

Matching the data is important, since I want to only capture the effect of a credit rating change on debt issuance and no other effects on debt issuance. For this reason I will use quarterly data. However, this has as a drawback that there will be less data available. An example of this is that most of the stockholder’s equity data from before the financial crisis is missing. While this isn’t a problem with the first two models, this is a serious concern with the third model. To make sure the results aren’t biased I will use yearly data to estimate the third model. The data on the credit ratings as well as the capital structure will come from Compustat. The credit ratings I use in the dataset are the S&P domestic long-term issuer ratings, which are the common ratings to use when researching the CS-CR hypothesis (see e.g. Kisgen, 2006 and Kemper and Rao, 2013). I use this credit rating because it is seen as the corporate credit rating which indicates the overall ability to pay its financial obligations.

IV. Results

This section will be used to discuss the results from estimating the models of the previous chapter. Table I contains summary statistics for the data used, including Net Debt Issuance, size, profitability, leverage, (in)tangibility and depreciation. On average, firms reduce net debt relative to total assets with 2.04% per quarter. The skewness supports this result since the distribution of the Net Debt Issuance is skewed to the left. This is consistent with less access to and overall reduction of debt since the financial crisis hit (see for example Gorton, 2010). The standard deviation of the Net Debt Issuance is large however, making it difficult to significantly interpret this. Furthermore this table gives the descriptive statistics of the control variables, which holds no surprise in values. All the observation are around 8,000 after dropping quarters with missing values or Net Debt issuance over 1 or under -1 (as previously done in similar studies like Kisgen, 2006 and Li et al., 2013). In Appendix B and Appendix C the detailed summary statistics are found for both the base sample as well as the expanded dataset, including firms of the S&P 400 MidCap index. The results shown for the

(17)

17 Table I

Summary Statistics S&P500

Mean, median, standard deviation, skewness and observations for quarterly data of S&P500 firms. Net Debt

Issuance is defined as the change of long-term debt and debt in current liabilities minus the change in

stockholder’s equity divided by total assets. Size is a control variable defined as the natural logarithm of sales. Profitability is a control variable defined as EBITDA divided by total assets. D/A is a control variable for the debt/assets ratio, defined as the long-term debt and debt in current liabilities divided by total assets. Intangibles is an extra control variable defined as intangible assets divided by total assets. Depreciation is an extra control variable defined as depreciation and amortization expense divided by total assets.

Mean Median Std. Dev. Skewness Observations Net Debt Issuance -0.0204 -0.0097 0.1083 -3.5716 8,097

Size 7.7378 7.5984 1.2561 0.4099 8,095

Profitability 0.0424 0.0388 0.0266 0.0818 7,958

D/A 0.2424 0.2236 0.1541 0.7151 8,097

Intangibles 0.2229 0.1800 0.1994 0.7667 8,097

Depreciation 0.0105 0.0090 0.0076 6.0527 7,953

expanded dataset are similar to the ones from table 1. The average Net Debt Issuance is somewhat larger for the expanded dataset, which indicates that the S&P400 MidCap firms reduced net debt more than S&P500 companies. This difference isn’t very large however. Appendix D includes the correlation matrix of the variables used in the different models. The correlations found here aren’t very high, which can be an indication of the absence of

multicollinearity.

Next I will discuss the results from the first regression of the previous section.

𝑁𝑒𝑡𝐷𝐼𝑠𝑠 = 𝛽0+ 𝛽1𝐶𝑅 + 𝛽2𝐶 + 𝜀

This model estimates the effect of the current credit rating on the issuance of debt. For every credit rating there is a specific dummy to measure if it has a significant effect. The results are shown in table II. There is an interesting parabola visible: top-rated firms mostly have negative Net Debt Issuance, middle-rated firms have positive Net Debt Issuance and low-rated firms again have negative Net Debt Issuance. For low rated firms this is intuitive, they are likely to be in financial distress and try to cut back on debt, or they have trouble refinancing their debt. For high credit ratings the cause of this negative result is somewhat ambiguous. Their high credit ratings means that they could more easily attract debt, but it might be the case that managers are afraid to lose their high rating and thus are wary to issue more debt. The results on the high and low ends of the credit ratings aren’t significant

(18)

18 Table II

Leverage changes per credit rating

This table gives coefficients and t-statistics for pooled time series regressions of quarterly net debt issuance (NetDIss) on credit ratings and control variables. The credit ratings used are the S&P domestic long-term issuer rating from years 2002-2014. AAA until CC are dummies indicating credit ratings. Size is a control variable defined as the natural logarithm of sales. Profitability is a control variable defined as EBITDA divided by total assets. D/A is a control variable for the debt/assets ratio, defined as the long-term debt and debt in current liabilities divided by total assets. *, **, *** indicate statistical significance at a 10%, 5% and 1% level respectively.

S&P 500 S&P 500 + S&P 400

AAA -0.011 (-0.99) -0.012 (-1.06) AA+ -0.017 (-0.96) -0.012 (-0.64) AA -0.004 (-0.35) -0.002 (-0.24) AA- 0.005 (0.52) 0.008 (0.93) A+ -0.0001 (-0.02) -0.002 (-0.34) A 0.003 (0.54) 0.004 (0.92) A- 0.013** (2.44) 0.11** (2.17) BBB+ 0.014*** (2.75) 0.015*** (3.42) BBB 0.010** (2.00) 0.012*** (3.20) BBB- 0.007 (1.34) 0.009** (2.32) BB+ 0.008 (1.17) 0.007 (1.53) BB -0.003 (-0.33) 0.010** (2.23) BB- -0.002 (-0.20) 0.004 (0.73) B+ -0.026* (-1.71) -0.004 (-0.50) B -0.003 (-0.17) 0.0002 (0.02) B- -0.121*** (-2.99) -0.036** (-2.20) CCC+ 0.022 (1.01) 0.021 (1.05) CC -0.092 (-0.86) 0.053 (0.64) Size 0.006*** (5.04) 0.005*** (4.77) Profitability -0.17*** (-3.67) -0.384*** (-10.51) D/A 0.107*** (12.28) 0.071*** (13.20) Intercept -0.091*** (-10.07) -0.063*** (13.20) R² 0.0357 0.0325 N 7,958 13,934

(19)

19 however, probably because the sample size is relatively small around those ratings. The B- rating is an exception to this, since it’s negative as expected and also significant. The significant results around the middle part of the credit rating distribution are another interesting finding, since they seem to indicate that firms with a middle-rating issue more net debt than firms with other ratings. There is a clear distinction to be made, namely the distinction between investment-grade firms and speculative-grade firms. Firms above a BB+ rating are investment-grade, firms below BBB- are speculative-grade. This distinction is important, since certain investors (banks, institutional investors) aren’t allowed to invest in speculative-grade bonds. This causes a lack of access to capital for firms with these kinds of bonds. The distinction is also visible in the results in table II. Firms with a credit rating just above the investment-grade rating tend to significantly issue more net debt. This can be an indication of the fact that these firms have easier access to capital than the firms just below the investment-speculative rating distinction. The R² of this regression is relatively low, but this isn’t a concern since it’s similar of size compared to research on the same subject. Table II also shows the result of this model for the combined dataset of the S&P500 and the S&P MidCap 400. The results are very similar, although the Net Debt Issuance on the lower end of the credit rating distribution isn’t negative as before, but insignificantly positive and close to zero.

The results of the second model, which takes into account the effect of a credit rating change, are shown in table III.

𝑁𝑒𝑡𝐷𝐼𝑠𝑠 = 𝛾0+ 𝛾1𝐶𝑅𝑑𝑜𝑤𝑛 + 𝛾2𝐶𝑅𝑢𝑝 + 𝛾3𝐶 + 𝜀

Regression (1) is the base model above. The coefficients of a credit rating upgrade or downgrade have the expected signs, an upgrade leads to more issuance of debt whereas a downgrade leads to less debt issuance. It is however surprising that the upgrade coefficient is significant while the coefficient indicating a downgrade is not. Based on previous literature the opposite would be expected. A possible explanation for this is found in the intercept. This intercept is negative and significant, unlike the ones found in similar research (e.g. Kemper and Rao, 2013). This might be a sign that overall debt has been reduced in the sample period. Given the economic circumstances firms had to deal with in that period, this

(20)

20 Table III

The effect of credit rating changes on leverage of the S&P500

This table gives coefficients and t-statistics for pooled time series regressions of quarterly net debt issuance

(NetDIss) on credit ratings and control variables. The credit ratings used are the S&P domestic long-term issuer

rating from years 2002-2014. Upgrade is a dummy variable equal to 1 if a credit rating is at least one fraction higher than in the previous period. Downgrade is a dummy variable equal to 1 if a credit rating is at least one fraction lower than in the previous period. Size is a control variable defined as the natural logarithm of sales.

Profitability is a control variable defined as EBITDA divided by total assets. D/A is a control variable for the

debt/assets ratio, defined as the long-term debt and debt in current liabilities divided by total assets. 10-year

Treasury rate is an extra control variable defined as the interest rate on a Treasury bond of the US with the

maturity of 10 years in percentages. Intangible is an extra control variable defined as intangible assets divided by total assets. Depreciation is an extra control variable defined as depreciation and amortization expense divided by total assets. Pre-Crisis Dummy is a dummy equal to 1 if the data comes from before January 1, 2008.

Industry Effects refers to including dummy variables for each industry. *, **, *** indicate statistical significance

at a 10%, 5% and 1% level respectively.

S&P 500 (1) (2) (3) (4) (5) Upgrade 0.019*** (2.73) 0.019*** (2.62) 0.019*** 0.019*** (2.73) 0.019*** (2.73) Downgrade -0.013 (-1.55) -0.013 (-1.54) -0.013 -0.014* (-1.67) -0.016* (-1.89) Size 0.005*** (5.76) 0.005*** (5.51) 0.005*** 0.006*** (5.76) 0.009*** (6.40) Profitability -0.181*** (-4.01) -0.170*** (-3.78) -0.153*** -0.192*** (-4.15) -0.299*** (-5.40) D/A 0.110*** (14.15) 0.111*** (14.30) 0.108 0.111*** (14.17) 0.158*** (13.38) 10-year Treasury rate - - -0.017*** - - Intangible - - - -0.002 (-0.37) - Depreciation - - - 0.158 (0.94) -

Pre-Crisis Dummy? No Yes No No No

Industry Effects? No No No No Yes

Intercept -0.081*** (-10.16) -0.079*** (-4.24) -0.031*** -0.083*** (-9.66) -0.077** (-2.35) R² 0.0318 0.0340 0.0464 0.0319 0.0484 N 7,958 7,958 7,921 7,951 7,958

isn’t surprising. Since upgrades were relatively rare during the financial crisis, it might be the case that a firm that experiences an upgrade gains relatively many benefits of debt so that this has caused the increased effect of an upgrade. The economic climate can also be the

(21)

21 reason why a downgrade doesn’t have a significant effect. As mentioned before,

downgrades have been common over the last decade, even when this doesn’t reflect a change in fundamentals. Managers might not act on this change as much as before, since the downgrades don’t reflect a deterioration of their ability to pay their obligations. The coefficients of the control variables all have the expected signs and are statistically

significant. Again, the R² of this regression is relatively low, similar to previous studies on this subject.

Regression (2) shows the effects of including a pre-crisis dummy to the base model. The quarterly data sample from 2002-2014 contains many missing values for the period before the financial crisis of 2008. To make sure the data from after the crisis isn’t biased by the few observations from before the crisis, the pre-crisis dummy is included. As can be seen in table III, the coefficients largely remain the same. An extra control variable, the 10-year US Treasury rate, is added in regression (3). Since the interest rate has been an instrument for the Federal Reserve to stimulate the economy in the sample period, as previously discussed in the literature section, this might have had an impact on the issuance of debt. Previous literature on the CS-CR hypothesis mainly ignore the Treasury rate as a control variable, but Kemper and Rao (2013) mention that including it has no effect on their results. To maintain comparability with other literature they drop this control variable. Table III shows that the same result is found here, in that the coefficients remain largely unchanged when adding the 10-year Treasury rate. The control variable itself has the expected sign, since a higher

Treasury rate will cause interest rates on corporate bonds to rise as well, which has a

negative effect on bond issuance. For the same reason as Kemper and Rao (2013) I drop this control variable in further regressions.

In regression (4), extra control variables are added. Intangibility is a variable that measures the tangibility of a firm’s assets. The reason this might influence a firm’s capital structure is that tangible assets are easier to value. This then gives greater security to lenders. Rajan and Zingales (1995) investigated this empirically and confirmed this positive effect. The

intangible coefficient is thus expected to be negative. In regression (4) this is indeed the case, but the coefficient is very low and insignificant. Another extra control variable added in regression (4) is depreciation. The reason this is added is because depreciation is a measure

(22)

22 Table IV

The effect of credit rating changes on leverage of the S&P500 and S&P400

rating from years 2002-2014. Upgrade is a dummy variable equal to 1 if a credit rating is at least one fraction higher than in the previous period. Downgrade is a dummy variable equal to 1 if a credit rating is at least one fraction lower than in the previous period. Size is a control variable defined as the natural logarithm of sales.

Profitability is a control variable defined as EBITDA divided by total assets. D/A is a control variable for the

debt/assets ratio, defined as the long-term debt and debt in current liabilities divided by total assets. 10-year

Treasury rate is an extra control variable defined as the interest rate on a Treasury bond of the US with the

maturity of 10 years in percentages. Intangible is an extra control variable defined as intangible assets divided by total assets. Depreciation is an extra control variable defined as depreciation and amortization expense divided by total assets. Pre-Crisis Dummy is a dummy equal to 1 if the data comes from before January 1, 2008.

Industry Effects refers to including dummy variables for each industry. *, **, *** indicate statistical significance

at a 10%, 5% and 1% level respectively.

S&P500 + S&P400 (1) (2) (3) (4) (5) Upgrade 0.015** (2.35) 0.015** (2.28) 0.016** 0.015** (2.35) 0.015** (-1.88) Downgrade -0.012 (-1.58) -0.011 (-1.57) -0.012 -0.013* (-1.71) -0.015* (-1.88) Size 0.005*** (7.36) 0.005*** (7.26) 0.005*** 0.005*** (7.42) 0.008*** (8.15) Profitability -0.416*** (-11.47) -0.399*** (-11.27) -0.389*** -0.425*** (-11.70) -0.529*** (-13.00) D/A 0.076*** (15.28) 0.077*** (15.34) 0.076*** 0.077*** (15.23) 0.086*** (14.00) 10-year Treasury rate - - -0.017*** - - Intangible - - - 0.002 (0.44) - Depreciation - - - 0.266** (2.01) -

Pre-Crisis Dummy? No Yes No No No

Industry effects? No No No No Yes

Intercept -0.062*** (-11.38) -0.061*** (-4.09) -0.014** -0.065*** (-10.86) -0.488*** (-9.23) R² 0.0303 0.0315 0.0412 0.0307 0.0478 N 13,934 13,934 13,868 13924 13,934

of fixed assets. More fixed assets lead to easier access to capital, thus the coefficient of depreciation is expected to be positive. This is indeed the case, while this effect is

(23)

23 coefficients don’t change much. The effect of a downgrade is now significant though, in contrast to the previous specifications. The coefficient is larger than the one Kisgen (2009) found using the sample period 1987-2003, which might indicate that the effect of a

downgrade has become larger over time. This will be separately tested and discussed later.

Industry variables are added in regression (5) of table III. Industry matters when determining capital structure, since certain industries experience different circumstances or are more subject to business cycles than others. Different industries might need different ways of financing, and access to capital might also differ between industries. These are all reasons to believe that this is an important factor to include. The coefficients of regression (5) don’t change much compared to the base model, and all coefficients are significant. To further check robustness of the results, table IV includes results for the extended sample including the S&P MidCap 400. These results are very similar to the main sample, concluding that the results don’t just hold for large firms.

The following that was investigated is the effect of an upgrade or a downgrade to

investment-grade and speculative-grade respectively. As mentioned previously, this is an important distinction with concern to the access to capital. The following regression was estimated:

𝑁𝑒𝑡𝐷𝐼𝑠𝑠 = 𝛾0+ 𝛾1𝐼𝑛𝑣𝐶𝑅𝑑𝑜𝑤𝑛 + 𝛾2𝐼𝑛𝑣𝐶𝑅𝑢𝑝 + 𝛾3𝐶 + 𝜀

where NetDiss and the control variables C are defined as before, and InvCRDown and InvCRup are dummies indicating a downgrade from investment-grade to speculative-grade and an upgrade from speculative-grade to investment-grade respectively. The idea behind this regression is that these credit rating changes are of larger importance than regular credit rating changes. The results for both samples are shown in table V.

The coefficients for an upgrade as well as a downgrade have the expected effect, which was also the case with the regular credit rating changes. However, in contrast to the previous found results, an upgrade now isn’t statistically significant for both samples. A downgrade to speculative-grade is statistically significant and about five times as large as a regular

(24)

24 Table V

The effect of a change in credit rating class on leverage for the S&P500 and S&P400

rating from years 2002-2014. InvUpgrade is a dummy variable equal to 1 if a credit rating is upgraded to investment-grade compared to the previous period. InvDowngrade is a dummy variable equal to 1 if a credit rating is downgraded to speculative-grade compared to the previous period. Size is a control variable defined as the natural logarithm of sales. Profitability is a control variable defined as EBITDA divided by total assets.

D/A is a control variable for the debt/assets ratio, defined as the long-term debt and debt in current liabilities

divided by total assets. *, **, *** indicate statistical significance at a 10%, 5% and 1% level respectively.

S&P500 S&P500 + S&P400

InvUpgrade 0.021 (0.89) 0.015 (0.73) InvDowngrade -0.081*** (-2.73) -0.123*** (-5.50) Size 0.005*** (5.68) 0.005*** (7.39) Profitability -0.182*** (-4.04) -0.407*** (-11.52) D/A 0.110*** (14.10) 0.077*** (15.41) Intercept -0.081*** (-10.03) -0.062*** (-11.36) R² 0.0316 0.0318 N 7,958 13,934

downgrades to speculative-grade, whereas the effect of an upgrade is quite similar across credit ratings. The effect of a downgrade might not be the result of management choices to issue less net debt to achieve a future upgrade, but just the effect of the inability to receive debt finance. This is consistent with the effect Kemper and Rao (2013) found.

Table VI gives the results from the third regression of section III, including interaction effects to capture the changed effects of upgrades and downgrades since the beginning of the financial crisis.

𝑁𝑒𝑡𝐷𝐼𝑠𝑠 = 𝜃0+ 𝜃1𝐶𝑅𝑑𝑜𝑤𝑛 + 𝜃2𝐶𝑅𝑢𝑝 + 𝜃3𝐶𝑅𝑑𝑜𝑤𝑛 ∗ 𝑃𝑟𝑒𝐹𝐶 + 𝜃4𝐶𝑅𝑢𝑝 ∗ 𝑃𝑟𝑒𝐹𝐶 + 𝜃5𝐶 + 𝜀

This was done with annual data, since there was a lack of quarterly data from before the start of the financial crisis. A drawback of using annual data is that it can create a mismatch, since over a year time there might be a lot of different things that influence capital structure besides credit rating changes. To make sure the sample is large enough, the extended

(25)

25 Table VI

The changed credit rating effect since the financial crisis

This table gives coefficients and t-statistics for pooled time series regressions of annual net debt issuance

rating from years 2002-2014. Upgrade is a dummy variable equal to 1 if a credit rating is at least one fraction higher than in the previous period. Downgrade is a dummy variable equal to 1 if a credit rating is at least one fraction lower than in the previous period. Upgrade*PreCrisis is a dummy variable equal to 1 if a credit rating is at least one fraction higher than in the previous period and takes place before January 1, 2008.

Downgrade*PreCrisis is a dummy variable equal to 1 if a credit rating is at least one fraction lower than in the

previous period and takes place before January 1, 2008. Not shown are control variables for Size, Profitability and D/A. The results for these control variables are similar to previous found results. *, **, *** indicate statistical significance at a 10%, 5% and 1% level respectively.

S&P500 + S&P400 Upgrade -0.009 (-0.95) Downgrade -0.025** (-2.43) Upgrade*PreCrisis 0.007 (0.49) Downgrade*PreCrisis 0.034** (2.29) Intercept -0.038 (-0.34) R² 0.1336 N 3,596

sample including the S&P MidCap 400 is used. This probably won’t bias the results, since previous regression made clear that the extended sample leads to similar results as the original sample.

The results of Table VI are similar for the effect of a downgrade compared to the quarterly sample, which is negative and significant. An upgrade has an insignificant negative effect, as predicted by Kisgen (2009), but in contrasts to the previous quarterly sample. This might be caused by the extension of the sample period to 6 years before the start of the financial crisis. The coefficients of interest in this regression however are Upgrade*PreCrisis and Downgrade*PreCrisis. These coefficients indicate if the effect of a credit rating change was different before the start of the crisis compared to after the start of the crisis. I expected both effects to become smaller, since credit ratings have become less informative over time. While this is insignificantly the case with an upgrade, the effect of a downgrade has

significantly become larger. So it is thus the case that not only the amount of debt outstanding has overall decreased, a credit rating downgrade also has a larger negative

(26)

26 effect on the issuance of debt since the start of financial crisis. A reason for this might be that the financial crisis has caused people to be more wary of credit ratings, and a

downgrade is viewed more negatively than before. Downgraded firms might receive even less favorable credit terms after the crisis than before, which causes them to cut on debt. Another reason might be that unwarranted downgrades causes managers to reduce debt because they try to get back to their original credit rating where they actually belong based on fundamentals.

The final question that will be investigated in this paper is the question if debt maturity also changes when a firm experiences a credit rating downgrade or upgrade. The model used to check this is the following:

𝑁𝑒𝑡𝑆ℎ𝑜𝑟𝑡𝐷𝐼𝑠𝑠 = 𝜑₀+ 𝜑₁𝐶𝑅𝑑𝑜𝑤𝑛 + 𝜑₂𝐶𝑅𝑢𝑝 + 𝜑₃𝐶 + 𝜀

Table VII gives the results from estimating this equation. Specification (1) is the base model as shown above. What can be seen is that a credit rating upgrade has a positive effect on the issuances of Net Short-term Debt, while a credit rating downgrade has the opposite effect. This is consistent with the findings of Bali and Skinner (2006) since firms with higher credit ratings issue debt with shorter maturity. The theory behind this is that these firms have positive future outlooks on credit conditions and that by the time of refinancing these conditions won’t be much worse. Firms with lower credit ratings are more uncertain about their future and thus don’t want to take the risk of early refinancing with worse credit terms.

Regression (2) adds a pre-crisis dummy to check whether these effects are different for the period before and after the start of the financial crisis. The coefficients remain largely the same, indicating that this effect remains fairly constant over the sample period. In regression (3) industry effects are added. These possibly have an effect since certain industries need different maturity than others as studied by Erhemjamts et al. (2010). Results in table VII show that while the R² increases by adding industry effects, the coefficients remain largely unchanged. The fourth specification (4) excludes speculative graded firms from the data sample. These firms are known to be in lesser financial conditions than firms with an

(27)

27

Table VII

The effect of a credit rating change on the maturity of issued debt

This table gives coefficients and t-statistics for pooled time series regressions of annual net short-term debt issuance (NetShortDIss) on credit ratings and control variables. The credit ratings used are the S&P domestic long-term issuer rating from years 2002-2014. Upgrade is a dummy variable equal to 1 if a credit rating is at least one fraction higher than in the previous period. Downgrade is a dummy variable equal to 1 if a credit rating is at least one fraction lower than in the previous period. Size is a control variable defined as the natural logarithm of sales. Profitability is a control variable defined as EBITDA divided by total assets. D/A is a control variable for the debt/assets ratio, defined as the long-term debt and debt in current liabilities divided by total assets. Pre-Crisis Dummy is a dummy equal to 1 if the data comes from before January 1, 2008. Industry Effects refers to including dummy variables for each industry. Investment-grade only excludes observations where a firm is speculative graded. *, **, *** indicate statistical significance at a 10%, 5% and 1% level respectively.

(1) (2) (3) (4) Upgrade 0.016*** (2.99) 0.016*** (3.07) 0.016*** (2.81) 0.014** (2.17) Downgrade -0.010* (-1.78) -0.010* (-1.79) -0.013** (-2.21) -0.017*** (-2.75) Size 0.003*** (2.84) 0.003*** (3.02) 0.002 (1.36) 0.002 (1.50) Profitability -0.041*** (-2.75) 0.042*** (-2.76) 0.057*** (-2.95) -0.067*** (-2.76) D/A -0.112*** (-13.36) -0.112*** (13.36) -0.172*** (-15.14) -0.116*** (-9.12)

Pre-Crisis dummy? No Yes No No

Industry Effects? No No Yes No

Investment-grade only? No No No Yes Intercept -0.009 (-1.13) -0.012 (-1.52) 0.068 (0.87) 0.0001 (0.01) R² 0.0527 0.0533 0.1207 0.0586 N 3,647 3,647 3,647 1,942

investment-grade credit rating. For this reason they might not be able to get access to long-term debt, something that Diamond (1991) argues as well. This fact might bias the previous results in that a downgrade for a speculative graded firm doesn’t cause them to issue more long-term debt, since they don’t have access to this kind of debt. The results in table VII for regression (4) indicate that this indeed is the case, because the downgrade effect is now larger than in previous specifications. Other coefficients are similar to the ones previously found.

(28)

28 V. Conclusion

This paper looks at the relationship between credit ratings and capital structure, more specifically the effect of a change in credit rating. What I find is that the effect of an upgrade is positive on the issuance of net debt, and a downgrade is negative. This effect holds after certain robustness checks including other control variables and different samples. These results partly tie in with previous findings on the CS-CR hypothesis such as Kisgen (2009), where the downgrade effect was also found. The effect of an upgrade is more surprising, since previous literature like Kisgen (2006) and Kemper and Rao (2013) argued that upgrades have a lesser effect than downgrades. A cause for the effect I find might be that upgrades have been relatively rare since the start of the financial crisis. This may have caused the benefits of a credit rating upgrade to become larger than before. An implication of these results is that current capital structure theories should be extended with a credit rating factor, since this has an influence on capital structure given that traditional capital structure factors are held equal.

The effect of a downgrade tends to be larger when the downgrade causes a firm to get a speculative-grade credit rating, which is consistent with the argument of Kisgen (2006) that costs of such a downgrade are larger. I don’t find an increased effect for an upgrade to investment-grade however. This might indicate that the downgrade effect mostly holds for these kinds of downgrades, while upgrades hold for all sorts of upgrades. Future research might further investigate this by looking at the effect of a credit rating change per different level of credit rating. A possible caveat here might be data availability, since very low graded firms are rare.

Another finding is that the previously mentioned effect has changed over time. This isn’t surprising, given that the whole economic climate has changed dramatically over the past decade. The effect of a downgrade has become larger than before, which was found in both the quarterly data sample as the annual sample. This is surprising since previous research found that credit ratings have become less informative over time (e.g. Baghai et al., 2013), so an opposite effect would be expected. A possible explanation for the findings in this paper is that the view of issuing debt has become more negative since the financial crisis,

(29)

29 which can be seen in the way that the Net Debt Issuance has been negative overall. This might also have caused managers to act more on a downgrade than before, or have caused lenders to ask less favorable credit terms which reduces access to capital.

Finally, in this research is investigated what the effect of a credit rating change is on the maturity of the debt issued. Previous research like Diamond (1991) and Bali and Skinner (2006) found that higher rated firms tend to issue more short-term debt than long-term debt. The findings of this paper support their research, since I find that an upgrade leads to more issuance of short-term relative to long-term debt while I find the opposite effect for a downgrade. When excluding speculative graded firms, this effect is found to be even larger.

Interesting to see is how these results will change in the future. With the financial crisis still in mind, the public opinion on debt is negative, which also has its effect on credit ratings. If this changes, it might be the case that the relationship between credit rating and capital structure also changes. The overall effect of a credit rating change will probably still be the same, the magnitude however is susceptible to change, as shown in this study.

A potential problem of this study might be the external validity. I only investigated data from American firms, it is unknown if the CS-CR relation still holds in other countries. While it might be difficult to gain information on credit ratings of different countries and gather a large enough sample size, this is potentially a subject for future research. Other possible extensions to this research relates to the fact that capital structure is more than debt alone. Equity is also a part of capital structure, and it might be the case that the market value of equity will decrease because of a credit rating downgrade. This is another way in which credit ratings can affect capital structure, and might be a subject for future research.

(30)

30 References

Altman, E.I., & Rijken, H.A. (2004). How rating agencies achieve rating stability. Journal of Banking and Finance, 28(11), 2679-2714.

Baghai, R.P., Servaes, H. & Tamayo, A. (2014). Have Rating Agencies Become More Conservative? Implications for Capital Structure and Debt Pricing. The Journal of Finance, 69(5), 1961-2005.

Bali, G. & Skinner, F.S. (2006). The Original Maturity of Corporate Bonds: The Influence of Credit Rating, Asset Maturity, Security, and Macroeconomic Conditions. The Financial Review, 41, 187-203.

Becker, B. & Milbourn, T. (2011). How did increased competition affect credit ratings? Journal of Financial Economics, 101(3), 493-514.

Begley, T. (2014). The Real Costs of Corporate Credit Ratings. Unpublished Paper, Retrieved from http://dx.doi.org/10.2139/ssrn.2404290

Blume, M.E., Lim, F. & Craig Mackinlay, A. (1998). The Declining Credit Quality of U.S. Corporate Debt: Myth or Reality? The Journal of Finance, 53(4), 1389-1413. Boot, A.W.A., Milbourn, T.T. & Schmeits, A. (2006). Credit Ratings as Coordination

Mechanisms. The Review of Financial Studies, 19(1), 81-118.

Bosch, O. & Steffen, S. (2011). On syndicate composition, corporate structure and the certification effect of credit ratings. Journal of Banking and Finance, 35(2), 290-299. Brounen, D., De Jong, A. & Koedijk, K. (2006). Capital structure policies in Europe: Survey

evidence. Journal of Banking and Finance, 30(5), 1409-1442.

Cheng, M. & Neamtiu, M. (2009). An empirical analysis of changes in credit rating properties: Timeliness, accuracy and volatility. Journal of Accounting and Economics, 47(1-2), 108-130.

Diamond, D.W. (1991). Debt Maturity Structure and Liquidity Risk. The Quarterly Journal of Economics, 106(3), 709-737.

Dichev, I.D. & Piotroski, J.D. (2001). The Long-Run Stock Returns Following Bond Ratings Changes. The Journal of Finance, 56(1), 173-203.

Dimitrov, V., Palia, D. & Tang, L. (2015). Impact of the Dodd-Frank act on credit ratings. Journal of Financial Economics, 115(3), 505-520.

(31)

31 Duca, M.L., Nicoletti, G. & Vidal Martinez, A. (2014). Global Corporate Bond Issuance What

Role for US Quantitative Easing? European Central Bank, 1649, 1-37.

Erhemjamts, O., Raman, K. & Shahrur, H. (2010). Industry Structure and Corporate Debt Maturity. The Financial Review, 45, 627-657.

Fama, E.F. & French, K.R. (2002). Testing Trade-Off and Pecking Order Predictions About Dividends and Debt. The Review of Financial Studies, 15(1), 1-33.

Faulkender, M. & Petersen, M.A. (2006). Does the Source of Capital Affect Capital Structure? The Review of Financial Studies, 19(1), 45-79.

Flannery, M.J. & Rangan, K.P. (2006). Partial adjustment toward target capital structures. Journal of Financial Economics, 79(3), 469-506.

Frank, M.Z. & Goyal, V.K. (2009). Capital Structure Decisions: Which Factors Are Reliably Important? Financial Management, 38(1), 1-37.

Gorton, G. (2010). Slapped by the Invisible Hand: The Panic of 2007. Oxford University Press, USA.

Graham, J.R. & Harvey, C.R. (2001). The theory and practice of corporate finance: evidence from the field. Journal of Financial Economics, 60(2-3), 187-243.

Harford, J., Klasa, S. & Walcott, N. (2009). Do firms have leverage targets? Evidence from acquisitions. Journal of Financial Economics, 93(1), 1-14.

Hite, G. & Warga, A. (1997). The Effect of Bond-Rating Changes on Bond Price Performance. Financial Analysts Journal, 53(3), 35-51.

Holthausen, R.W. & Leftwich, R.W. (1986). The effect of bond rating changes on common stock prices. Journal of Financial Economics, 17(1), 57-89.

Jensen, M.C. & Meckling, W.H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3(4), 305-360.

Joyce, M.A.S., Lasoasa, A., Stevens, I. & Tong, M. (2011). The Financial Market Impact of Quantitative Easing in the United Kingdom. International Journal of Central Banking, 7(3), 113-161.

Kayo, E.K. & Kimura, H. (2011). Hierarchical determinants of capital structure. Journal of Banking and Finance, 35(2), 358-371.

Kemper, K.J. & Rao, R.P. (2013). Do Credit Ratings Really Affect Capital Structure? Financial Review, 48(4), 573-595.

Credit Ratings and Capital Structure : what is the relationship and has this changed over time?